55
meter. Leaf blades were oven dried at 40°C for 10 days (Heraeus, Germany) to measure main culm LDW (mg). LMA per main culm (mg/mm2) was calculated at stage x as:
56
Six biological replicates per accession were randomly hand harvested to collect biological yield (BY; determined by weighing the total air-dried aboveground parts). Single-plant grain yield and yield components were measured by counting the number of grains per main spike, tillers and spikes number per plant. Main spike dry weight MSDW and total grain weight per main spike after hand threshing were measured. Harvest index (HI) per plant was measured as the ratio of grain weight per plant to BY per plant multiplied by one hundred.
3.3.4 Data analyses
Growing conditions were arranged in a completely randomized design with three replicates and each growing condition includes border plants which were not sampled. Analysis of variance (ANOVA) was conducted to compare spike row-type classes between growing conditions, and to compare row-type across growing conditions using SAS for Windows ver.
9.3 (SAS Institute Inc., Cary, NC, USA) at P≤0.05). Means were separated according to the Fisher’s Least Significant Difference (LSD) at 0.05 levels of probability. Pearson’s correlation coefficient was calculated for phenotyped traits using PROC CORR (SAS, 2013) at P≤0.05. Variance components were estimated by considering the genotype x growing condition. Broad-sense heritability (H2) was estimated overall growing conditions according to Snedecor and Cochran (1980) using PROC VARCOMP (SAS, 2013)
Equation 3-3: Broad-sense heritability (H2)
H
2= σ2g(σ2g +σ2g × gw/e +σ2e/re)
where σ2 g is genotypes variance; σ2g.gw is the variance of the interaction genotype × growing condition, r is replicates, and e the error.
Principal Component Analysis (PCA) based on phenotypic correlations of accession means values for each selected trait under each growth condition was calculated using GENSTAT for Windows version 16 (VSN International, Hemel Hempstead, UK). A two-dimensional PCA was calculated as multivariate analysis to interpret and summaries phenotypic
57
clusters/variations among growth conditions and/or row-type classes of barley by accessions by LA, LDW, and LMA at different growth stages (GenStat, 2013). PCA is an indicator ordination tool for obtaining multivariate data that can be explored visually in a two-dimensional PCA correlation. Coefficients of variation (CV) for individual trait were calculated as a percentage of standard deviation to the trait mean by GenStat (2013). CV was calculated to compare individual leaf trait variations within growth conditions between accessions.
PCA analysis identified phenotypic groups within two- and six-rowed barley collections and/or growing conditions by using LA, LDW and LMA data at the HD stage. A two-dimensional scatter plot is presented to show the row-type clusters based on leaf traits (Figure 3-1A), the clusters of row-type based on growing conditions (Figure 3-1B) and the clusters of row-type within growing conditions (Figure 3-1C). The first PC-1 (LA, LDW or LMA) accounted for 48.97% of phenotypic variation and clearly separated two- and six-rowed accessions with few exceptions (Figure 3-1A). LA and LDW had the major contributions in this separation and the mixed dots (accessions) between row-types are from six-rowed barleys in the GH. The second PC-2 (growing conditions) accounted for 34.14% of the observed phenotypic variation (Figure 3-1A). Accessions from different growing conditions were mixed within each row-type and cannot be clustered based on leaf traits only, indicating
intra-Yield components
Thermal time/Growing degree-day (GDD)
Growing condition
Awn Primordium Tipping Heading Anther Extrusion
Two-rowed
Six- rowed
Two-rowed
Six-rowed
Two-rowed
Six-rowed
Two-rowed
Six-rowed Leaf number
per main culm
GH 0.34 0.69 -0.21 -0.27 -0.27 -0.20 0.06 -0.17 Pots 0.10 -0.23 0.45 0.46 0.56 0.45 -0.19 0.56
Soil 0.07 0.20 0.65 0.48 0.63 0.48 0.35 0.42
Leaf fresh weight per main culm
(g, LFW)
GH 0.67 0.68 0.70 0.64 0.36 0.32 0.36 0.18
Pots 0.70 0.59 0.69 0.75 0.52 0.77 0.22 0.71
Soil 0.40 0.47 0.49 0.74 0.43 0.24 0.32 0.23
Leaf dry weight per main culm
(mg, LDW)
GH 0.52 0.65 0.68 0.55 0.68 0.48 0.65 0.46
Pots 0.76 0.81 0.75 0.62 0.55 0.77 0.30 0.76
Soil 0.52 0.49 0.46 0.70 0.71 0.48 0.39 0.29
Leaf area per main culm (mm,
LA)
GH 0.40 0.37 0.23 -0.22 0.30 0.20 0.42 -0.09
Pots 0.59 0.60 0.67 0.79 0.40 0.84 0.43 0.75
Soil 0.40 0.27 0.52 0.70 0.53 0.48 0.72 0.42
Leaf mass area per main culm (mg.mm-2, LMA)
GH 0.30 0.25 0.28 -0.32 -0.32 -0.24 -0.24 -0.42 Pots 0.49 0.72 0.28 0.11 0.31 -0.01 0.82 0.40 Soil 0.46 0.50 0.05 -0.22 -0.11 0.20 -0.47 0.02
59
class variation, while growing conditions explained leaf traits variation within each row-type (Figure 3-1B and -C). Herein, we cannot rule out the effect of germplasm diversity (geographical origins) and/or germplasm status (cultivar, landrace and line) on phenotypic diversity.
To compare individual leaf trait variations within growing conditions between accessions, we examined coefficients of variation (CV, Supplementary Table 3-1). For all measured traits, CV values under GH conditions were always higher than those of field-grown plants (pots and soil). Generally, CV values increased for all measured traits after AP under all growing conditions. LA under GH conditions resulted in intermediate to very high CV values over developmental stages, CV from 27% to 80% and 28% to 41% in two- and six-rowed barley, respectively. Low to intermediate CV values over developmental stages in field-grown plants (pot and soil) ranged from 14% to 35% and 16% to 31% in two- and six-rowed barley, respectively. Other leaf traits followed the trend, clearly suggesting that GH conditions (i.e.
controlled temperature, light etc.) maximized phenotypic leaf trait expression between accessions compared to field conditions. The obtained and high CV values for our GH conditions indicate that they are appropriate for studying phenotypic/genetic variation of leaf traits.
The analysis of variance components among row-type classes and growing conditions at developmental stages highlighted that the genetic variance is the largest components (Figure 3-2A). Importantly, we found the broad-sense heritability values across growing conditions ranged from high to very high (H2 = 0.66-0.90) at all developmental stages (Figure 3-2B).
These values, most notably, suggest that leaf traits in both row-type classes are mainly heritable under all growing conditions at all developmental stages.
60
Figure 3-1: Principal component analysis (PCA) based on phenotypic data by leaf area (LA), leaf dry weight (LDW) and leaf mass area (LMA) per main culm at heading stage.
PCA for row-types, green and blue color denotes two- and six-rowed barley, respectively (a), and for growth conditions, black, greenish-brown and gray color denotes for greenhouse (GH), pot- and soil-field, respectively (b), and row-types within growth conditions (c); the light-red and dark-blue circles indicate two- and six-rowed plants, respectively, under GH conditions; the green and light green circles indicate two- and six-rowed plants, respectively, under pot-field conditions; and the gray and dark-red circles indicate two- and six-rowed plants, respectively, under soil-field conditions. The number of dots (n = 54, 18 x 3 growth conditions) for six-rowed and (n = 42, 14 x 3 growth conditions) for two-rowed barley.
61
Figure 3-2: Variance components of leaf traits at different developmental stages in two- and six-rowed barley together overall growing conditions (A). Broad-sense heritability of leaf traits at different developmental stages overall growing conditions in two- and six-rowed barley separately (B).
3.4.3 Leaf traits in both row-type classes under different growth conditions at four
62
exhibited the largest LA, followed by pot-grown field plants and GH-grown plants (independent of row-type) (Supplementary Figure 1B). LDW followed the LA trend for all development stages, growth conditions, and row-type classes (Figure 3C and Supplementary Figure 1C). The LMA was significantly different between two- and six-rowed barley (P ≤ 0.05) at AP, TIP, and HD stages independent of growth conditions, but not at AE (Figure 3D). Field-grown plants consistently showed significantly higher LA and LDW than GH-grown plants. However, GH conditions produced the broadest phenotypic variation for leaf traits between accessions, which is important for revealing the genetic variation of these traits. Interestingly, we found that six-rowed barley plants had greater LA and LDW, but lower LMA compared to two-rowed barley plants under all growth conditions. The found natural variation for LA and related traits between row-type classes suggests that the effect of major genes controlling row-type in barley may have pleiotropic effects on these traits.
However, additional genetic evidence is required to elucidate this effect.
63
Figure 3-3: Leaf traits per main culm at different developmental stages for two- and six-rowed barley averaged across all growth conditions.
A) Leaf number; B) leaf area; C) leaf dry weight; and D) leaf mass area. The same letters at each developmental stage are not significantly different at P ≤ 0.05 according to the Least Significant Difference. Bars indicate standard deviation [n = 42 (14 x 3) and 54 (18 x 3) for two- and six-rowed barley, respectively]. AP: awn primordium, Alqudah and Schnurbusch (2014); TIP: tipping, Z49; HD: heading, Z55; AE: anther extrusion, Z65, Zadoks et al. (1974)
64
3.4.4 Leaf growth rate in both row-type classes under different growing conditions at four
65
Figure 3-4: Leaf growth rate per main culm at different stages and phases for two- and six-rowed barley averaged across all growth conditions.
A) Leaf fresh weight growth rate (LFWGR); B) leaf dry weight growth rate (LDWGR); C) leaf area growth rate (LAGR); and D) leaf dry matter content (LDMC). The same letters at each developmental stage are not significantly different at P ≤ 0.05 according to the Least Significant Difference. Bars indicate standard deviation [n = 42 (14 x 3) and 54 (18 x 3) for two- and six-rowed barley, respectively]. AP: awn primordium, Alqudah and Schnurbusch (2014); TIP: tipping, Z49; HD: heading, Z55; AE: anther extrusion, Z65, Zadoks et al. (1974). 2R is two-rowed and 6R is six-rowed.
66
3.4.5 Correlation analysis between leaf area (LA), single plant yield, and yield components Correlation coefficients at developmental stages between LA (i.e. LA of the main culm) and main culm spike grain yield, main spike dry weight (MSDW), tillers and spikes per plant, BY, and HI were generally higher under field conditions (pots and soil; Table 3-4). Correlations at HD and AE stages were higher than for AP and TIP stages (Table 3-4). In both row-type classes under all growing conditions, we report positive correlations between LA per main culm with main culm spike grain yield (grain number and weight) and MSDW at all developmental stages (Table 3-4). The strongest correlations were found in field-grown six-rowed barleys at the AE stage. For HI, we always found a negative correlation with LA in two-rowed barley in all growing conditions (Table 3-4). For tiller number, spike number per plant, and BY, correlation analysis showed two-rowed barley at AE stage under GH conditions exhibited the strongest correlation between LA with tiller and spike number per plant (r = 0.61 and 0.51, respectively; Table 3-4), while LA from two-rowed barley in field-grown plants (pot and soil) highly correlated with BY at the AE stage (r = 0.65). Unlike in two-rowed barley, six-rowed barley showed consistent positive association between LA of the main culm and HI, whereas LA in six-rowed barley correlated negatively with BY, tiller and spike number per plant under all growing conditions. The correlation results indicate that two-rowed types produced more BY, tillers and spikes per plant; whereas six-two-rowed barleys produced more grains (per spike). In general, larger LA in six-rowed barley correlated with main spike grain yield, MSDW, and HI, while smaller LA in two-rowed barley influenced tiller and spike number per plant and BY. LA for field-grown plants (pots and soil) correlated better with yield and yield components compared with those grown in GH conditions.
This study focused on the importance of variation between two- and six-rowed barley in LA and its relationship on single plant grain yield and yield components. The study also demonstrates that six-rowed barley had larger and heavier leaf blades compared to two-rowed barley. Moreover, we identified AP to TIP as being the most critical sub-phase for leaf growth and development. We also investigated the heritability of leaf traits (i.e. genetic basis ) in barley row-type classes under various growth conditions and throughout developmental stages.
Yield components
Leaf blade area per main culm (mm2) Growing
condition
Awn
Primordium Tipping Heading Anther Extrusion
Two-rowed
Six- rowed
Two-rowed
Six-rowed
Two-rowed
Six-rowed
Two-rowed
Six-rowed Grain number
per main culm spike at
harvest
GH 0.13 0.20 0.14 0.28 0.37* 0.52* 0.43* 0.62*
Pots 0.57* 0.55* 0.45* 0.56* 0.53* 0.64* 0.60* 0.74*
Soil 0.21 0.38* 0.51* 0.61* 0.53* 0.62* 0.58* 0.65*
Grain weight per main culm spike at
harvest (g)
GH 0.12 0.13 0.26 0.30 0.23 0.40* 0.33 0.46*
Pots 0.53* 0.40* 0.40* 0.49* 0.35* 0.50* 0.50* 0.55*
Soil 0.28 0.34 0.42* 0.62* 0.47* 0.53* 0.55* 0.68*
Main culm spike dry weight at heading (g)
GH 0.16 0.10 0.35* 0.34 0.31 0.24 0.33 0.36*
Pots 0.10 0.36* 0.37* 0.79* 0.43* 0.57* 0.65* 0.83*
Soil 0.31 0.10 0.56* 0.69* 0.43* 0.54* 0.74* 0.74*
Tillers per plant
GH 0.31 -0.07 0.01 -0.17 0.47* -0.33 0.61* -0.36*
Pots 0.23 -0.26 0.20 -0.10 0.15 0.00 0.45* -0.09 Soil 0.03 -0.17 0.28 -0.24 0.39* -0.38* 0.45* -0.31 Spikes per
plant
GH 0.25 -0.06 0.30 -0.54* 0.43* -0.34 0.51* -0.27 Pots 0.22 -0.30 0.20 -0.23 0.38* -0.22 0.47* 0.00 Soil 0.16 -0.22 0.29 -0.22 0.33 -0.40* 0.40* -0.31 Biological yield
(g)
GH 0.00 -0.38* 0.27 -0.32 0.29 -0.21 0.39* -0.11
Pots 0.46* -0.08 0.47* -0.07 0.36* -0.13 0.47* -0.31 Soil 0.20 -0.29 0.42* -0.29 0.52* -0.27 0.65* -0.52*
Harvest Index (%)
GH -0.26 0.38* -0.14 0.37* -0.29 0.12 -0.17 0.20
Pots -0.35* 0.29 -0.22 0.16 -0.39* 0.21 -0.17 0.28 Soil -0.22 -0.06 -0.12 0.22 -0.31 0.54* -0.30 0.41*
68
3.5.1 The importance of leaf area in improving single-plant yield
In this study, we found significant variation in LA and related traits during growth and development stages between barley row-types. Using LA of single leaf (flag or penultimate) for improving the yield of barley has been previously suggested (Berdahl et al., 1972;Rasmusson, 1987); however, no study has explored the specific contribution of main culm LA during pre-anthesis stages to yield based upon row-type classes. Our analysis showed that larger LA (six-rowed) correlate to higher main culm spike grain yield, spike dry weight and harvest index; while smaller LA (two-rowed) correlated to more spikes, tillers, and biological yield. We attempt to clarify the importance of large and small LA in improving yield which may help breeders in future breeding programs.
3.5.1.1 Large leaf area
In this study we showed that larger LA per main culm in six-rowed barley positively associates with a higher grain yield per main spike (Table 3-4), thereby increasing HI. LA is known to improve grain yield in breeding programs (Richards, 1983;Rasmusson, 1987;Peng et al., 2008). Broader leaves were important components in New Plant Type (NPT) breeding program to improve rice grain yields (Khush, 1995;Fujita et al., 2013). Rasmusson (1987) suggested that larger LA produces larger spikes and kernels, and in two bi-parental barley populations (small vs. large LA), higher grain weights and higher yield was achieved by higher photosynthetic rates per unit LA (Berdahl et al., 1972). A similar explanation for higher single plant grain yield in six-rowed barley is likely to be the cause in this study. Here, a high correlation between larger LA per main culm in six-rowed barley with MSDW at the AE (flowering) stage was apparent. Yoshida (1972) found a close relationship between LA at flowering time and grain number and grain yield at harvest. We similarly propose that larger LA leads to more dry matter accumulation before the AE stage from higher photosynthetic rates, which in turn increases spike dry matter. This could be one reason for producing heavier
69
single spike grain weight in six-rowed barley and higher HI. As presented in Figure 3, larger LA at the HD stage resulted in the highest amount of LDW in six-rowed barley, which is a result that may be attributed to the previous reasoning (high photosynthetic rate in large LA) and confirmed in past literature (Aase, 1978). Our observation highlights the importance of large LA in six-rowed barley, but differs from the ideotype concept of Donald (1968).
Therefore, based on our observed correlations, improving single-plant grain yield through larger LA in six-rowed barley might be important for future barley breeding programs in the context of increasing spike grain yield, spike dry weight, and harvest index.
3.5.1.2 Small leaf area
We found that smaller LA in two-rowed barley was associated with more tillers and spikes per plant, which are findings that are in agreement with those of Berdahl et al. (1972).
Producing smaller and narrower LA in dense stands can theoretically improve crops grain yield in cereals (Donald, 1968). As shown in Figures 3-3 and 3-4, smaller LA in two-rowed barley results in higher LMA and LDMC, suggesting that the leaves are thicker than those of six-rowed barley. LMA is a trait that responds to stress, as it reflects the amount of dry matter a plant accumulates through reduced LA (Witkowski and Lamont, 1991). Smaller LA, which results from a longer vegetative duration, could be one important trait for improving drought and cold tolerance depending on growth habits (Van Oosterom and Acevedo, 1993). Curtis et al. (2012) reported that leaves with higher LMA were thicker, narrower, and protected well against rapid fluctuations in temperature, while leaves with low LMA senesced earlier. Thus, two-rowed barley may adapt better to stress conditions than six-rowed barley based on leaf performance traits, such as LMA. Although LMA strongly varied under different conditions, such as water stress (Poorter et al., 2009), it may still be a useful trait to select for in a breeding program in order to produces lines with improved stress tolerance. Moreover, in a canopy situation, narrower leaves utilize sunlight more efficiently due to decreased shading between tillers and neighboring plants, thereby increasing light perception. Taken together,
70
the opposing relationship between LA and yield-related traits, such as BY and HI, in two- and six-rowed barley, respectively, is undoubtedly an interesting plant architectural feature and deserves further attention.
3.5.2 Maximized leaf growth rate in the two row-type classes
Our study provides a first set of leaf growth rate parameters at four developmental stages and in three growing conditions for two major barley row-type classes. Previous research focused on the importance of the leaf appearance/emergence rate (phyllochron) in barley and wheat instead of leaf growth rate parameters (Kirby et al., 1985;Kirby, 1995). An initial lag phase up to AP stage existed for all parameters studied (Figure 3-5). Leaf weight and LA significantly increased with the onset of the late reproductive phase and reached a plateau at HD stage (Figure 3-5). Increased leaf weight and area resulted from a higher number of developed leaves during more advanced developmental stages. As presented in Table 3-1, LA is critical for biomass accumulation processes (LDW) and is indicative of increased photosynthetic rates (Berdahl et al., 1972). This, in turn, may explain increased LFWGR and LDMGR. Moreover, we identified the period from AP to TIP as the most important sub-phase for leaf growth rate in barley (Figure 3-3). This period is characterized by 50% LA and biomass accumulation, but decreases with advancing age (after HD) due to leaf senescence. This concords with Alqudah and Schnurbusch (2014), who found that the AP to TIP sub-phase is the most critical period for spikelet survival with approximately ~70% of total spikelet/floret abortion.
Importantly, the AP to TIP period temporally coincides with stem elongation and maximal leaf growth rates, further corroborating that within-plant competition (stem/leave vs. spike) may be a major trigger for observed spkikelet/floret decline (Alqudah and Schnurbusch 2013).
These findings illuminate opportunities to better understand internal networks of competing barley organs and might open up novel research in terms of source-sink relationships (Sadras and Denison, 2009).
71
Figure 3-5: General trend of leaf traits in barley at developmental stages.
Data was collected from nine biological replicates of 32 barley accessions (n=32 accession x 3 reps x 3 growth conditions).
3.5.3 Genetic background and variance analysis
One major objective was to estimate variance components and broad-sense heritability (H2) values for different leaf traits under variable growth conditions. Interestingly, all heritability values across growth conditions obtained for leaf traits were above 0.65 in small LA (two-rowed) and large LA (six-(two-rowed), suggesting that these traits are predominantly genetically controlled (Figure 4-2). Rebetzke et al. (2004) reported that heritability values for estimated LA per plant in wheat (30 genotypes) and barley (3 genotypes) at an early stage (4 leaves stage) of development under different conditions (GH and field) was approximately 0.90. In a study conducted by Rasmusson (1987), LA heritability only ranged from 0.24-0.37 in three barley populations. One reason for the discrepancy between our findings and those of Rasmusson (1987) may be that we evaluated leaf traits in a diverse collection (different regions and genetic background) under two environmental conditions (GH and field) and at different developmental stages. Evidently the genetic component is the largest component in our diverse collection, and since environmental effects are small, genetic regulation plays a major role in the observed variation. Moreover, it is possible that single plant leaf measurements are more reliable than leaf measurements conducted in dense stands (i.e., canopy or plot situation) (Fasoula and Fasoula, 1996). For example, LA components in alfalfa
72
plants significantly differ as a direct result from stand density and light competition (Baldissera et al., 2014). Similar factors may be in play for barley plants grown under light competition in a field-plot situation, thus creating variable environmental factors (light competition and shading) that cause shoot and leaf growth variation. Therefore, estimating the genetic components of leaf traits is more accessible under controlled conditions using single plants (Fasoula and Fasoula, 1996).
Our results show that measured leaf traits are mainly genetically determined and less affected by environment. For instance, LMA was affected by many factors (within species or/and environments) for which a large part of the variation was still unaccounted (Poorter et al., 2009). In the literature, the heritability value for LMA is low and is considered as a complex trait, and understanding the genetic factors underlying this trait has been unsuccessful to date.
In the present study, we found that LMA was predominantly genetically controlled (H2 ≥ 0.68), which was likely due to our analysis of this trait under different environments in a diverse collection and at different developmental stages (i.e., experimental design). We also examined whether growth conditions and/or row-type classes affected leaf traits. Leaf traits were generally higher for field-grown plants than for GH-grown plants, which might be due to greater space for root growth and nutrient availability. Results from the present study showed that leaf traits are influenced by growth conditions to a greater extent than phase duration or spike-related traits (Alqudah and Schnurbusch, 2014). In the GH, we tried to minimize environmental effects by randomizing pots several times a week; yet, we still postulate that field conditions at several locations/seasons are important for further leaf trait validation work. Certainly, leaf traits are similarly affected by pot size and/or substrate content in the GH, but CV values under GH conditions were always higher than those of field-grown plants clearly suggesting that single-plants in the GH maximized phenotypic leaf trait expression between accessions. and therefore are more appropriate for studying phenotypic/genetic variation of leaf traits (Fasoula and Fasoula, 1996). Based upon the PCA analysis, leaf traits
73
(PC-1, 48.97%) were major distinguishing features of row-type classes. However, we cannot rule out the effect of germplasm diversity (geographical origins) on phenotypic variation.
In this study, we found significant differences in leaf traits between two- and six-rowed barley, especially LA. Phenotypic differences in LA between row-type classes showing a high heritability value provide an unexplored opportunity to better understand the genetic of LA in barley. Whether the phenotypic differences between row-type classes in LA and other traits may be related to the action of the predominant row-type gene SIX-ROWED SPIKE 1 (Vrs1) is not yet clear. Future research is required to elucidate these relationships in more detail.
74
4.0 CHAPTER FOUR: Genetic Dissection of Photoperiod Response Based on GWAS of Pre-Anthesis Phase Duration in Spring Barley
Ahmad M. Alqudah1, Rajiv Sharma2,3, Raj K Pasam2, 4, Andreas Graner2, Benjamin Kilian2, 5 and Thorsten Schnurbusch1*
1Research group Plant Architecture, 2 Research group Genome Diversity, Genebank Department, Leibniz Institute of Plant Genetics and Crop Plant Research (IPK), Corrensstrasse 3, OT Gatersleben, D-06466 Stadt Seeland, Germany
3Present address: College of Life Science, University of Dundee at The James Hutton Institute, Invergowrie, Dundee DD2 5DA, UK
4Present address: Biosciences Research Division, Agriculture Productivity, Department of Environment and Primary Industries, Agribio 5, Ring rd, Bundoora, Victoria 3083, Australia
5Present address: Bayer CropScience NV, Innovation Center, BCS R&D–Trait Research, Technologiepark 38, 9052 Zwijnaarde (Gent), Belgium
*Corresponding author, email: thor@ipk-gatersleben.de
PLOS.ORG , USA
PLOS ONE, 9, e113120, DOI:10.1371/journal.pone.0113120
Received 24 Jun2014, accepted 23 October 2014, Published online November 24, 2014
4.1 Abstract
Heading time is a complex trait, and natural variation in photoperiod responses is a major factor controlling time to heading, adaptation and grain yield. In barley, previous heading time studies have been mainly conducted under field conditions to measure total days to heading. We followed a novel approach and studied the natural variation of time to heading in a world-wide spring barley collection (218 accessions), comprising of 95 photoperiod-sensitive (Ppd-H1) and 123 accessions with reduced photoperiod sensitivity (ppd-H1) to long-day (LD) through dissecting pre-anthesis development into four major stages and sub-phases.
The study was conducted under greenhouse (GH) conditions (LD; 16/8 h; ~20/ ~16 oC day/night). Genotyping was performed using a genome-wide high density 9K single nucleotide polymorphisms (SNPs) chip which assayed 7842 SNPs. We used the barley physical map to identify candidate genes underlying genome-wide association scans (GWAS).
GWAS for pre-anthesis stages/sub-phases in each photoperiod group provided great power for partitioning genetic effects on floral initiation and heading time. In addition to major genes
Heading time is an important trait for barley (Hordeum vulgare L.) adapting to particular environmental cues and hence for maximizing grain yield. Its complex genetic architecture was considered as one of the major breeding goals during last century. The capacity to regulate heading time provides crop plants with the opportunity to successfully complete their life cycle under a wide range of environments, which exceed the distribution range of their wild relatives (Andres and Coupland, 2012). The optimal time to flower is crucial for high crop grain yields (Purugganan and Fuller, 2009). For decades, studies on heading/flowering time solely focused on the total number of days until heading/flowering and its effect on grain yield in response to environmental cues. However, the time prior to anthesis (i.e. pre-anthesis developmental phases) in barley consists of vegetative, early and late reproductive phases (Sreenivasulu and Schnurbusch, 2012;Alqudah and Schnurbusch, 2014). Grain yield and yield potential are significantly influenced by the reproductive, pre-anthesis phase durations (Borràs et al., 2009), which were shown to be genetically controlled (Alqudah and Schnurbusch, 2014). Despite the importance of this earlier phase of development, most of what we know about the genetic control of pre-anthesis phases is purely based on traditional quantitative trait locus (QTL) analysis, wherein often vegetative and reproductive phases were not clearly separable (Borràs-Gelonch et al., 2010;Borràs-Gelonch et al., 2011). Alqudah and
76
Schnurbusch (2014) proposed an amended approach for dissecting the longest pre-anthesis phase (late-reproductive phase) into three sub-phases: awn primordium (AP) to tipping (TIP);
TIP to heading (HD); HD to anther extrusion (AE). This refined approach, based upon clearly defined developmental pre-anthesis stages, might shed more light on the causal genetic factors responsible for the variation in developmental stages/sub-phases in response to photoperiod in barley.
Most studies in barley aimed to unravel the genetics of heading time and the underlying specific genes in response to photoperiod, vernalization and/or earliness per se. The first two factors change heading time in response to environmental conditions, while the last factor determines heading time independent of photoperiod and temperature (Bullrich et al., 2002).
In barley, a long-day (LD) crop, the PSEUDO-RESPONSE REGULATOR (HvPRR37) gene, also known as PHOTOPERIOD RESPONSE LOCUS1 (Ppd-H1), is the central heading time gene regulated in responses to LD, at which recessive alleles (ppd-H1) reduce the response to LD (Turner et al., 2005). Variation at Ppd-H1 affects heading time of accessions originating from different geographical regions. Spring barley accessions originating from Middle East, e.g. tend to carry photoperiod responsive Ppd‑H1 alleles, causing early heading under LD, while the delay of heading time in Northern European accessions of spring barley is due to reduced photoperiod sensitivity, ppd-H1, alleles (Turner et al., 2005;Andres and Coupland, 2012). Therefore, strength of photoperiod response is a key factor to understand the natural genetic variation and pathway of heading time in barley.
In addition to the HvPRR37 gene, which is located on the short arm of chromosome 2H (Turner et al., 2005), further genes of the heading time pathway have been identified in barley. Ppd-H2, responsive to short-day (SD) is located on 1HL, for which HvFT3 has been proposed as a candidate gene (Faure et al., 2007). Five FLOWERING LOCUS T (FT-like;
HvFT1-HvFT5) genes were found in barley, and these genes play various roles during plant development through their photoperiod response, of which HvFT1 has a major role in the
77
transition from the vegetative-to-reproductive phase as an important source of variation in heading time (Faure et al., 2007). Moreover, the CONSTANS-like (CO) gene family is known to regulate flowering time through the photoperiod pathway in Arabidopsis (a LD plant) and rice (SD plant). In barley,Griffiths et al. (2003);Cockram et al. (2012) identified numerous homologs of CO-like genes (HvCO1 to HvCO18) but their roles in the barley heading time pathway are still unclear. GIGANTEA (GI)-CO-FT is considered as a conserved central interaction partner in plant photoperiod pathway under LD, in Arabidopsis (Higgins et al., 2010); however, the function of HvGI in the barley photoperiod pathway is still unclear. CCT domain gene families (CO, CO-LIKE, TIMING OF CAB1 (TOC1)), i.e. CO-like and PRR, have an important role in controlling heading time; in addition to these families (Cockram et al., 2012) introduced uncharacterized genes carrying single CCT domains called CCT MOTIF FAMILY (CMF) genes. PHYTOCHROME (HvPhy) and CIRCADIAN CLOCK ASSOCIATED (HvCCA) genes clearly affect barley heading time pathway through interaction with other genes, such as HvPhyC which induces early heading by up-regulating HvFT1 and bypassing HvCO1 under LD (Nishida et al., 2013).In Arabidopsis, LATE ELONGATED HYPOCOTYL (LHY) and CIRCADIAN CLOCK ASSOCIATED 1 (CCA1) suppress FT expression independent of the CO causing delayed flowering (Fujiwara et al., 2008). Similarly, SHORT VEGETATIVE PHASE-like (SVP-like) genes such as VEGETATIVE TO REPRODUCTIVE TRANSITION gene2 (HvVRT2) in barley delayed heading time by inhibiting spike development under LD (Trevaskis et al., 2007). RICE FLORICAULA/LEAFY (i.e. RFL; syn.
ABERRANT PANICLE ORGANIZATION2, APO2) is the homolog of Arabidopsis LEAFY and plays important roles in regulating the transition from vegetative to reproductive phase, maintenance of inflorescence meristem, floral organ identity/determinacy and flowering time in rice (Rao et al., 2008;Higgins et al., 2010;Ikeda-Kawakatsu et al., 2012). In Arabidopsis, LFY acts downstream of SUPPRESSOR OF OVEREXPRESSION OF CONSTANS 1, (SOC1) (Higgins et al., 2010); whereas RFL functions upstream of OsSOC1 and reduced expression of
78
RFL delayed flowering time in rice (Rao et al., 2008). With regard to genes that are involved in responses to vernalization, Vrn-H1 located on 5HL (promotes transition from the vegetative to the reproductive phase) is dominant in spring barley (Hemming et al., 2008), while Vrn-H2 (HvZCCT) located on 4HL delays heading in plants that have not been vernalized (Karsai et al., 2005); it similarly delays heading time under LD (Casao et al., 2011a). Vrn-H3 (syn.
HvFT1), a central integrator of different heading time pathways, had been identified on 7HS (Yan et al., 2006). In addition, many independent EARLINESS PER SE (EPS) and EARLY MATURITY (EAM) loci have been identified in barley: eps2S (eam6) on 2HS, eps3L (eam10) on 3HL, eps4L on 4HL, eps5L on 5HL, eps6L.1 and eps6L.2 on 6HL, eps7S on 7HS and eps7L on 7HL (Laurie et al., 1995). The precise position of these genes in coherent barley heading time pathway is not yet understood.
High-throughput genotyping platforms recently developed in barley provide sufficient marker coverage to perform genome-wide association scans (GWAS) (Waugh et al., 2009). GWAS is a powerful tool for mapping complex plant traits, with unprecedented genetic resolution for gene identification in large-genome crops such as barley and wheat. GWAS can identify genes responsible for natural phenotypic variation through screening a large, diverse collection of accessions with high density genetic markers to find causal genes as a result of historical recombination (Waugh et al., 2009). In barley, GWAS has been used to identify single nucleotide polymorphism (SNP) markers associated with heading time (Comadran et al., 2012;Pasam, 2012). However, information on the genetic variation of pre-anthesis stages/sub-phases as key components of barley adaptation and grain yield is still lacking.
The aim of this study is to detect QTL underlying natural variation of pre-anthesis stages/sub-phases based upon differences in photoperiod response (Ppd-H1/ppd-H1) through dissecting time to heading into sub-phases in a world-wide spring barley collection. To achieve this objective, we phenotyped more than 3,000 plants at four developmental stages (AP, TIP, HD and AE) under controlled GH conditions, derived from 95 photoperiod responsive (Ppd-H1)
79
and 123 accessions with reduced photoperiod sensitivity (i.e. ppd-H1), respectively.
Distinction of these two photoperiod groups in our GWAS analysis allowed us to control population structure, while using a 9K SNP chip provided us with an unprecedented genetic resolution for studying the natural variation of time to heading. In combination with accurate phenotyping of pre-anthesis stages into sub-phases (i.e. sowing to AP, includes vegetative and early reproductive phases; AP-TIP; TIP-HD and HD-AE) within each photoperiod group, natural genetic variation of the time to heading could be genetically dissected resulting in the identification of novel QTL that were anchored to the barley physical map (e.g. several associations around HvCO-like genes). Clearly, novel rich genomic regions with highly associated SNPs were detected, which have not been detected before. This paper proposes a new heading-time model for barley with specific reference to allelic combinations to photoperiod-response groups.
80
Table 4-1: Spike row-type and origins of spring barleys accessions with photoperiod-sensitive (Ppd-H1) and reduced photoperiod sensitivity (ppd-H1).
‡ WANA: West Asia and North Africa, EU: Europe, EA: East Asia, AM, Americas.
The collection was genotyped using the 9K iSelect SNP chip from Illumina, which was developed from RNA-seq data of 10 diverse barley cultivars (Comadran et al., 2012). Finally for our GWAS analyses, we focused on SNPs which had genetic and physical positions on the barley genome after quality control checking, filtering and evaluating 9K SNP (Comadran et al., 2012;Pasam et al., 2012;Mascher et al., 2013). In each group, only the SNPs with minor allele frequency (MAF) ≥ 5% were used for association analyses (4228 and 4050 SNPs for Ppd-H1and ppd-H1group, respectively). We used genetic marker positions anchored by physical map positions for SNP markers based on Barke x Morex RILs POPSEQ population (Mascher et al., 2013).
The population structure of this collection was determined by 6355 polymorphic SNPs (Figure 4-1). In this study, we divided this collection based on photoperiod response (Ppd-H1and ppd-H1) as major groups and geographical regions present as sub-groups. Principal component analysis (PCA) was also used to infer the population structure in this collection.
PCA is an indicator ordination tool for obtaining clusters, which can be explored visually in a two dimensional using GenStat (GenStat, 2014).
Origin‡
Photoperiod-sensitive (Ppd-H1)
Reduced photoperiod
sensitivity (ppd-H1) Total Two-rowed Six-rowed Two-rowed Six-rowed
WANA 12 21 11 1 45
EU 10 6 80 12 108
EA 0 28 2 6 36
AM 6 12 4 7 29
Total 28 67 97 26
95 123 218
Thirty-five seeds of each accession from the collection of 218 spring barley accessions were germinated for 10 days under controlled condition in GH (LD condition, 16/8 h day/night and
~20/ ~16 oC day/night). Seedlings were transferred to vernalization chamber (SD condition, 10/14 h and ~4 oC) for a period (28 days) when they reached 2-3 leaves stage. Afterwards seedlings were kept in an acclimation chamber for a period of 7 days (16/8 h and ~14 /~12
oC). Finally, strongest 30 seedlings of each accession were transplanted into 0.5 liter pots (one plant per pot; nine centimeter pot diameter and nine centimeter height) under GH condition.
Plants were grown in a substrate containing peatmoss with 14: 16: 18 / Nitrogen (N):
82
Phosphorous (P): Potassium (K). To avoid any mineral deficiency each pot was additionally fertilized with 1.5 gram of solid fertilizer (that constitute minerals 17:11:10 / N: P: K). Plants were irrigated daily and supplemental light (~300 mE m-2s-1 PAR) was used to extend the natural light with low intensity incandescent lamps. Pots were randomized three times per week to reduce border and temperature-gradient effects on plant growth and development.
The time for each stage was recorded when at least 50% of the main culm spikes in each accession had reached at the stage: awn primordium stage, AP (Z31-33, maximum yield potential; Alqudah and Schnurbusch (2014)); tipping stage, TIP (Z49, first awns visible on main culm); heading time stage, HD (Z55, half main culm spike emerged from flag leaf sheath); anther extrusion stage, AE (Z65, half of main culm spike with anthers; Zadoks et al.
(1974)); more information on the stages can be obtained in 4- 2 and (Alqudah and Schnurbusch, 2014). To identify the exact timing for the AP stage, regular dissection (three times per week) from immature barley main culm spikes was performed by microscopic dissection and image capture using Stereo Microscope Stemi 2000-C with KL 1500 LCD;
Axio Vision, 4.8.2, ZEISS, Germany.
83
Figure 4-2: General figure of barley pre-anthesis phases.
The figure includes the beginning developmental point from sowing to each stage (e.g. time to tipping is 1200 GDD; i.e. from sowing time to tipping stage) and the differences between stages (phase; e.g. tipping to heading phase (150 GDD); i.e. GDD for heading stage (1350 GDD) minus tipping stage (1200 GDD)). GDD is the average GDD in the whole collection (i.e. including both photoperiod groups). This figure also describes the developmental stages and sub-phases which form the late reproductive phase as described in (Alqudah and Schnurbusch, 2014).
Thermal time (oC.D-1) or growing degree-days (GDD, base temperature was 0 oC) from sowing to reach each stage was recorded to measure the required days/thermal temperature for each stage and the duration between the stages (Table 4-S1). The phenotypic analyses of pre-anthesis sub-phases (i.e. dissection work) had to be performed in eight batches due to limited GH space and feasibility of workload. Each batch contained thirty accessions plus two check lines (Morex and Barke); hence, all 218 accessions were grown from September 2011 to April 2012 in eight consecutive batches for dissection and phenotypic, stage-specific data collection under controlled GH conditions. Each batch of accessions had a completely randomized design with 30 plants per accession. At least three biological replications at each developmental stage were recorded for analysis. The collected data were analyzed using SAS software version 9.3 at probability level P≤0.05. Student’s t-test was used to compare between
84
Ppd-H1and ppd-H1 groups. REML (Residual Maximum Likelihood) was used to analyze phenotypic data by SAS software (SAS, 2013). Best Linear Unbiased Estimates (BLUEs) were used to estimate phenotypic means for each trait in individual accession and estimated means were used for association analysis. Broad-sense heritability for traits in each group was calculated across growing times as the ratio between the genetic variance and the phenotypic variance which includes genotypic by growing times (environment) interaction variance and error variance components using PROC VARCOMP (SAS, 2013).
85
threshold to determine highly associated QTL within confidence interval ±5 cM. Known heading time genes (bold and italicized in Figure 4-6 and 4-7) have been genetically anchored and located according to the Barke x Morex RILs (POPSEQ) sequence contigs;
more information about these genes, their accession numbers and genetic chromosome positions are available in Table 4-S2.
4.4 Results
4.4.1 Population structure of a world-wide spring barley collection
The population structure of this collection was determined using polymorphic SNP data from the 9K array. To this end, we divided the collection into two groups based on the presence of a single diagnostic SNP in HvPRR37, thereby separating photoperiod responsive (i.e.
photoperiod-sensitive, Ppd-H1) accessions from those with reduced photoperiod sensitivity (i.e. ppd-H1; Turner et al. (2005); Sharma et al. in preparation; Figure 4-1). The relationships among spring barley accessions were inferred using principle component analysis (PCA). The genetic variation between groups was explained by PCA and the collection was clearly separated into Ppd-H1 and ppd-H1 spring barleys based on heading time data (Figure 4-S1).
PCA-1 explained 28.1% of variation and separated the ppd-H1-group from the Ppd-H1-group, with few exceptions, clearly showing greater genetic variation among accessions from the Ppd-H1-group compared to ppd-H1-carrying accessions (Figure 4-S1). Interestingly, the genetic variation at heading time could be further subdivided based on geographic origins (Figure 4-S2). The European (EU) accessions clustered from the remaining regions with few exceptions (Figure 4-S2). Although a significant proportion of the collection clustered separately based on photoperiod response, row-type classes formed another determinant within the geographic and photoperiod groups. Notably, most of the accessions in the Ppd-H1-group are six-rowed barleys from West Asia and North Africa (WANA) and East Asia (EA), while most of the accessions in the ppd-H1-group are two-rowed barleys from EU (Table 4-1). Generally, these results suggest that the spring barley collection (218 accessions)
It is difficult to understand the full complexity of the time to heading in cereals by only studying the period from sowing until heading/flowering. Our analyses aimed at examining associations for particular pre-anthesis stages or sub-phases, to maximize the likelihood of finding new associations. Hence, we developmentally dissected the pre-anthesis time of barley into four stages (AP, TIP, HD and AE) and four sub-phases (sowing-AP, AP-TIP, TIP-HD and TIP-HD-AE; Figure 4-2). The first investigated pre-anthesis stage was AP. Plants at this stage had already passed the vegetative-to-reproductive transition and finished early spike differentiation. The time from sowing to AP represents approximately 40% of the entire time to AE in barley and on average it took 650 growing degree-days (GDD, Figure 4-2). Late reproductive development in barley can be further sub-divided into three sub-phases, of which AP-TIP is the longest phase with an average of 550 GDD (Figure 4-2).
Comparisons of thermal time to reach different developmental stages in Ppd-H1and ppd-H1 spring barley groups under GH condition yielded highly significant variation between the groups (P≤0.05; Figure 4-3 and Table 4-S1). The duration from sowing to reach each developmental stage was significantly longer in the ppd-H1group at all stages except AP stage; i.e. at TIP, HD and AE the differences were +302, +406 and +373 GDD, respectively, in favor of the ppd-H1 group (Figure 4-3). Separate analysis of thermal time to reach developmental stages between geographical regions in Ppd-H1 and ppd-H1 spring barley groups can also explain the genetic variation (Figure 4-4A and B). In the Ppd-H1 group, there was no clear trend of results between geographical regions (Figure 4-4A), whilst the genetic variation between geographical regions were much clearer in the ppd-H1 group (Figure 4-4B). In particular EU accessions of the ppd-H1 group had the longest durations
87
from sowing to different pre-anthesis stages (TIP, HD and AE stages) except AP stage (Figure 4-4B).
Durations between sub-phases can explain the significant variation between the two photoperiod groups. The duration between AP-TIP was the longest late-reproductive sub-phase in both groups resulting in 690 GDD for the ppd-H1-carrying accessions but only 400 GDD for the Ppd-H1 group (Figure 4-3). The duration between TIP-HD and HD-AE was not significantly different between both groups. Generally, the AP-TIP sub-phase is the most important developmental period related to the observed genetic variation for the time to heading between these groups. In terms of origin, EU accessions had the longest duration between AP-TIP and TIP-HD sub-phases in both groups (Figure 4-4A and B). The genetic variation in pre-anthesis stages between geographical regions became clearer in the ppd-H1 group. Thus, the results of our dissection of the pre-anthesis stages provide a promising route to explain the genetic variation between the photoperiod groups (Ppd-H1 and ppd-H1) in barley.
Notably, the broad-sense heritability values for pre-anthesis developmental stages and sub-phases in each photoperiod group (Ppd-H1and ppd-H1) were above 0.88, indicating that the traits related to pre-anthesis phase duration are highly heritable (Table 4-2). Due to the very high heritability values associated with pre-anthesis stages we are able to detect particular QTL for each stage and sub-phase within each photoperiod group.
88
Figure 4-3: Thermal time for different developmental stages and sub-phases in photoperiod-sensitive (Ppd-H1) and reduced photoperiod sensitivity (ppd-H1).
Thermal time from sowing to the beginning of awn primordium (AP), tipping (TIP), heading (HD) and anther extrusion (AE) stages and thermal time of the duration of sub-phases. Letters differentiate between photoperiod-sensitive (Ppd-H1) and reduced photoperiod sensitivity (ppd-H1). The same letters are not significantly different at P≤0.05. Bars indicate standard deviation (n = 95 and 123 for Ppd-H1 and ppd-H1 barleys, respectively).
Table 4-2: Estimation of broad-sense heritability (H2) for developmental stages and sub-phases measured as thermal time oC.D-1 (GDD) in the association mapping groups.
†AP:
awn primordium (Alqudah and Schnurbusch, 2014); TIP: tipping, Z49; HD: heading, Z55; AE: anther extrusion, Z65 (Zadoks et al., 1974).
H2: broad-sense heritability for each group overall growing times based on accessions mean. n = 95 and 123 for barleys with photoperiod-sensitive and reduced photoperiod sensitivity, respectively.
Developmental stages calculated based on thermal time oC.D-1 (GDD).
Stage/phase† Photoperiod-sensitive
(Ppd-H1)
Reduced photoperiod sensitivity (ppd-H1)
Awn primordium (AP) 0.92 0.90
Tipping (TIP) 0.91 0.92
Heading (HD) 0.90 0.92
Anther extrusion (AE) 0.91 0.93
AP-TIP 0.89 0.91
TIP-HD 0.92 0.90
HD-AE 0.90 0.88